logging-log4j-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Remko Popma <remko.po...@gmail.com>
Subject Re: Aw: Re: Re: RollingRandomAccessFile and emtpy log file
Date Wed, 03 Feb 2016 22:28:51 GMT
You mention that RollingRandomAccessFile appender with immediateFlush=false and a buffer size
of 256K did not flush to disk until you logged 600KB. Looking at the code I don't see how
this is possible but you may have found a bug. Is this reproducible? If so, can you raise
a Jira with a small test program that reproduces the issue?

Continuing our discussion about workarounds:

You can use either appender. The performance numbers (random access file appender being 1.3
- 2 times faster than file appender) are for immediateFlush=false. I don't have numbers for
immediateFlush=true. 

There's a separate attribute for setting the buffer size (for both appenders). Perhaps a smaller
buffer size would flush sooner, have you experimented with that?

Remko

Sent from my iPhone

> On 2016/02/04, at 0:06, Veit Guna <Veit.Guna@gmx.de> wrote:
> 
> So, If I would like to use a rolling file appender with synchronous logging, but _with_
buffering and so _not_ flushing on every event what should
> I use? RollingRandomAccessFile with immediateFlush="true" or the "normal" RollingFileAppender
:)? Does immediateFlush="true" take the buffer into account?
>  
> Thanks
> Veit
>  
> 
> Gesendet: Mittwoch, 03. Februar 2016 um 13:43 Uhr
> Von: "Remko Popma" <remko.popma@gmail.com>
> An: "Log4J Users List" <log4j-user@logging.apache.org>
> Betreff: Re: Aw: Re: RollingRandomAccessFile and emtpy log file
> Sorry my message was unclear.
> 
> RollingRandomAccessFile appender does not _need_ async loggers, it's just that if you
choose to use async loggers, you will get the side-effect of regular flushing in the background.
So it achieves your goal (of flushing the buffer to disk.)
> 
> You can also set immediateFlush to true to achieve this, but this will write to disk
on _every_ event *in your application thread*.
> 
> If performance is important, async logging is 10-20 times faster than synchronous logging.
The difference between the normal file appender and random access file is much smaller (random
access file appender is only 1.3 - 2 times faster than file appender).
> 
> :-)
> 
> Sent from my iPhone
> 
>> On 2016/02/03, at 21:01, Veit Guna <Veit.Guna@gmx.de> wrote:
>> 
>> immediateFlush="true" seems to work - without adding async config options. Adding
the system property does also work, but needs disruptor.jar on
>> the classpath - as it seems. Otherwise a ClassNotFoundException occurs. Although
the documentation states, that the RandomAccessFile appenders do _not_ need the disruptor.jar
;).
>> 
>> Ok then. Since I don't want to add an additional dependency, I would go for immediateFlush="true".
But now I'm wondering whether it would be better to just use the normal RollingFileAppender
with not specifying flush behavior at all. Which one is faster in synchronous mode :)?
>> 
>> Thanks
>> Veit
>> 
>> 
>> 
>> Gesendet: Mittwoch, 03. Februar 2016 um 12:37 Uhr
>> Von: "Veit Guna" <Veit.Guna@gmx.de>
>> An: log4j-user@logging.apache.org
>> Betreff: Aw: Re: RollingRandomAccessFile and emtpy log file
>> Hi Remko.
>> 
>> I've read about the AsyncLoggerContextSelector but I didn't try it - to be honest.
As the documentation doesn't state anything about
>> that the RollingRandomAccessFile appender needs async functionality, I didn't took
this into account.
>> Maybe there could be added some hint in the documentation about that (needs async
or not)? Or maybe just don't allow starting up log4j when
>> the configuration is (obviously) wrong :)?
>> 
>> So I'll check for the async configuration option and report back.
>> 
>> Thanks!
>> Veit
>> 
>> 
>> 
>> 
>> Gesendet: Mittwoch, 03. Februar 2016 um 12:13 Uhr
>> Von: "Remko Popma" <remko.popma@gmail.com>
>> An: "Log4J Users List" <log4j-user@logging.apache.org>
>> Betreff: Re: RollingRandomAccessFile and emtpy log file
>> I would say immediateFlush but that doesn't explain why you're not seeing the update
until after logging 600 KB when the buffer is only 256 KB...
>> 
>> Have you tried setting system property -DLog4jContextSelector=org.apache.logging.log4j.core.async.AsyncLoggerContextSelector
>> to switch on async loggers? This will ensure the buffer is always flushed to disk,
but it uses smart batching to achieve this in an efficient manner (and it happens on the background
thread, so no impact on your application).
>> 
>> Sent from my iPhone
>> 
>>> On 2016/02/03, at 17:47, Veit Guna <Veit.Guna@gmx.de> wrote:
>>> 
>>> Hi.
>>> 
>>> I'm using log4j 2.5 with the RollingRandomAccessFile. When the application starts
up and performs logging, the logfile
>>> gets created, but nothing is written to it. Only after I stop the application
the content appears. I also doublechecked the 256KB
>>> buffer. But logging 600KB doesn't seem to trigger a write to the file. Console
appender is working fine.
>>> 
>>> Any idea what I'm missing?
>>> 
>>> Here's my configuration:
>>> 
>>> --cut here--
>>> <?xml version="1.0" encoding="UTF-8"?>
>>> <Configuration>
>>> 
>>> <Properties>
>>> <Property name="logsPath">${sys:app.logs:-${env:APP_LOGS:-logs}}</Property>
>>> </Properties>
>>> 
>>> <Appenders>
>>> <RollingRandomAccessFile name="rollingFileAppender" fileName="${logsPath}/app.log"
filePattern="${logsPath}/app-%d{yyyy-MM-dd}-%i.log" immediateFlush="false">
>>> <PatternLayout>
>>> <pattern>%d %-5p [%mdc{RQID}] [%c{1}] - %m%n</pattern>
>>> </PatternLayout>
>>> <Policies>
>>> <SizeBasedTriggeringPolicy size="10 MB" />
>>> </Policies>
>>> <DefaultRolloverStrategy max="5" />
>>> </RollingRandomAccessFile>
>>> <Console name="consoleAppender" target="SYSTEM_OUT">
>>> <PatternLayout pattern="%d{HH:mm:ss,SSS} %-5p [%mdc{RQID}] [%c{1}] - %m%n"
/>
>>> </Console>
>>> </Appenders>
>>> 
>>> <Loggers>
>>> <Logger name="com.acme.app" level="info" />
>>> 
>>> <!-- spring -->
>>> <Logger name="org.springframework.security" level="info" />
>>> <Logger name="org.springframework.transaction" level="info" />
>>> <Logger name="org.springframework.orm.jpa" level="info" />
>>> 
>>> <!-- log SQL statements -->
>>> <Logger name="org.hibernate.SQL" level="error" />
>>> <!-- log bind parameters -->
>>> <Logger name="org.hibernate.type" level="error" />
>>> 
>>> <Root level="info">
>>> <AppenderRef ref="consoleAppender" />
>>> <AppenderRef ref="rollingFileAppender" />
>>> </Root>
>>> </Loggers>
>>> 
>>> </Configuration>
>>> --cut here--
>>> 
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: log4j-user-unsubscribe@logging.apache.org
>>> For additional commands, e-mail: log4j-user-help@logging.apache.org
>> 
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: log4j-user-unsubscribe@logging.apache.org
>> For additional commands, e-mail: log4j-user-help@logging.apache.org
>> 
>> 
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: log4j-user-unsubscribe@logging.apache.org
>> For additional commands, e-mail: log4j-user-help@logging.apache.org
>> 
>> 
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: log4j-user-unsubscribe@logging.apache.org
>> For additional commands, e-mail: log4j-user-help@logging.apache.org
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: log4j-user-unsubscribe@logging.apache.org
> For additional commands, e-mail: log4j-user-help@logging.apache.org
>  
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: log4j-user-unsubscribe@logging.apache.org
> For additional commands, e-mail: log4j-user-help@logging.apache.org
> 

---------------------------------------------------------------------
To unsubscribe, e-mail: log4j-user-unsubscribe@logging.apache.org
For additional commands, e-mail: log4j-user-help@logging.apache.org


Mime
View raw message